Automatic target recognition for radar images can be utilized for differentiating commercial and non-commercial targets (e.g., maritime vessels, aircraft, etc.); facilitating identification of targets on borders; and identifying friendly or foe targets. However, since there are many possible types of targets, it is challenging, if not impossible, to correctly identify all known targets. Further, previous attempts at target identification have utilized simulated data to train technology to identify targets. These previous attempts do not take into account issues with real-world radar images such as noise spikes, clutter returns, range and Doppler smearing, and atmospheric confrontations. Thus, a need exists in the art for improved target identification for a radar image.